The mean K-MMSE score of all 214 patients was 22.3, and 133 patients (62.1%) had a K-MMSE score ≤ 23. The decreased K-MMSE score was obvious in most pathologies, including unruptured aneurysm (61.7%), ruptured aneurysm (64.0%), ischemic disease (70.0%) and trauma (70.6%), although there were some differences according to their treatment modalities. The proportion of patients with a K-MMSE score ≤ 23 was the highest in trauma and ischemic disease and the lowest in epilepsy and tumor patients (27.3% and 50.0%, respectively) (Table 1).

A decreased K-MMSE score was prevalent in almost all age groups, regardless of gender. In age groups older than 39 years, over 60.0% of patients had a K-MMSE score ≤ 23. Inter-gender differences were not statistically significant in all age groups. Although the overall mean K-MMSE score was higher in males than in females (22.7 and 22.0, respectively), this difference was not statistically significant (p = 0.285, using t-test) (Table 2).

We classified 90 aneurysm patients into four groups; group A (unruptured aneurysm, clipping), group B (unruptured aneurysm, endovascular coiling), group C (ruptured aneurysm, clipping) and group D (ruptured aneurysm, endovascular coiling) (Table 3). The proportion of patients in these groups with a K-MMSE score ≤ 23 was 67.7%, 53.3%, 60.0% and 65.5%, respectively. However, there was no statistically significant difference when patients were divided by treatment modality (group A vs. B and C vs. D, p = 0.914 and 0.478, respectively). There was also no difference in K-MMSE score between patients with unruptured vs. ruptured aneurysms (group A vs. C and group B vs. D, p = 0.802 and 0.550, respectively).

From 59 patients (27.6%) who completed the K-MMSE more than once, the first and last scores were compared. Their mean time after treatment was 29.3 months, and the mean time interval between test administrations was 11.9 months. The number of patients with a K-MMSE score ≤ 23 was 41 and 18 (69.5% and 30.5%) for the first and the last test, respectively. The mean K-MMSE score difference was 2.7, which was statistically significant (p = 0.000, using paired t-test) (Table 4). In all, 45 of 59 patients (76.3%) had an improved K-MMSE score. These improvements were notable in unruptured and ruptured aneurysm groups (89.5% and 75.0%, respectively). However, patients with vascular disease, such as Moyamoya disease or carotid stenosis had decreased K-MMSE scores, although these changes were not statistically significant (p = 0.700, using Wilcoxon signed-ranks test) (Table 4). The linear correlation between the time interval and the K-MMSE score change was not definite in the unruptured or ruptured aneurysm groups (R² = 0.078 and 0.002, respectively).

The purpose of this cross-sectional study was to measure the cognitive function of general neurosurgical patients. To measure the effect of disease or treatment modality on cognitive function, a case-control or comparative study can be good designs to compare two sets of data obtained from cases and controls or from the same individual before and after a procedure. However, these study designs are not always powerful. For example, in slowly progressing diseases such as tumor, we do not always know its outbreak to obtain initial data to use as a baseline. In emergency conditions, such as cerebral hemorrhage or trauma, we are also unable to know when we should obtain initial baseline data. Therefore, we thought the cross-sectional study design was a good choice to overcome these difficulties.

The MMSE was originally developed to evaluate elderly psychiatric patients.6)10) Although it is the most widely used screening instrument for dementia or cognitive dysfunction, it has some limitations.7)9)10) First, it cannot be fully administered to disabled individuals with motor impairment because it contains performance tests that assess praxis and visuospatial function.10) Therefore, we conducted our study in an outpatient clinic setting to exclude patients who were either in an acute postoperative period or critical condition. We also excluded those who never had brain surgery or interventions to maintain homogeneity of our case population; most of them had either very mild symptoms or very critical problems that were inoperable, and we thought they might cause a bias and skew the results. Second, it is difficult to distinguish postoperative cognitive dysfunction from postoperative delirium using the MMSE.19)21) By conducting an outpatient clinic based-study, we excluded postoperative delirium, which tends to be a transient and fluctuating disturbance of consciousness that occurs shortly after surgery, whereas postoperative cognitive dysfunction is a more persistent problem of cognitive performance that can be assessed by neuropsychological tests.16) Our patients' mean time from treatment to study was 30.9 months, and over 60% (130 of 214) of patients participated more than 1 year after treatment. Third, the MMSE needs to be translated and adapted to the Korean cultural background.7) Because simply translating the tests into foreign languages would not provide comparative results, we needed to adopt a verified measuring instrument.7)13)

The MMSE was modified and translated into Korean by Kang et al., and the resulting K-MMSE has been widely used in clinical evaluations and research involving Korean dementia patients.7)9) We defined a K-MMSE score ≤ 23 as the cut-off value for abnormal cognitive function for two reasons. First, it was used as a cut-off value in many other articles.4)9)11)14)21) With the same cut-off value, we can share established knowledge with other studies and compare them with ours. Second, we thought 23 is reasonable cut-off value, considering the mean K-MMSE score measured by Han et al. in 977 normal elderly Koreans.7) Although there are some variations according to age, gender and education level, the mean K-MMSE score of each group ranged from 23.3 to 27.7.7) Setting the cut-off value was very critical because more than 20% of patients (49 out of 214) were clustered around a score of 24 in our series, and this number was equivalent to more than half of normal patients (Fig. 3). If the cut-off value was 24 instead of 23, the percentage of normal patients would decrease from 37.9% to 15.0%.

More than 62% of patients (133 out of 214) had K-MMSE scores ≤ 23, and this decrease was common in almost all pathologies and treatment modalities in our study (Table 1). After stratification according to gender and age, most groups still had an average K-MMSE score ≤ 23. There was no statistically significant difference between genders (Table 2). In other articles, cognitive dysfunction has been documented in up to 50% of patients who have experienced subarachnoid hemorrhage.11)12) Several studies gave incidence estimates for post-stroke dementia that range between 6 and 32%.2) Compared with other type of surgery, patients who undergo brain surgery seem to be more vulnerable to cognitive dysfunction. Abildstrom et al.1) reported 10.4% of patients who underwent major abdominal, non-cardiac thoracic, or orthopedic surgery with general anesthesia had cognitive dysfunction 1-2 year postoperatively. Moller et al.14) also demonstrated that 9.9% of patients had cognitive dysfunction 3 months postoperatively. Although it is difficult to directly compare our results with other articles because of different study populations and various measurement instruments, our proportion seems much higher than that reported for non-brain surgery. We believe this difference stems from effects on the brain itself, either by pathologies or by treatments. Some studies on general versus regional or spinal anesthesia have not found differences in postoperative cognitive dysfunction.4)16) We think postoperative cognitive dysfunction is affected more by the type of surgery than anesthesia. Cognitive impairment was documented in up to 50% of patients of cardiac surgery patients.8)15)23) This is because cardiac surgery is a lot similar to brain surgery in that cardiopulmonary bypass is thought to have cerebral effect.14)16)23)

To figure out if the K-MMSE score was influenced by different pathologies or treatment modalities, we analyzed the data collected from aneurysm patients (Table 3). Because both unruptured and ruptured aneurysms share the same treatment modalities, we could analyze the effect of pathologies when the effect of treatment modalities was controlled (group A vs. C and B vs. D). At the same time, we could analyze the effect of treatment modalities when the effect of pathology type was controlled (group A vs. B and C vs. D). Because ruptured aneurysms develop from unruptured aneurysms, we expected to estimate the additive effect of subarachnoid hemorrhage. However, we failed to find significant differences in each analysis. There was no difference between pathologies or treatment modalities. Considering the articles that report no cognitive function change after craniotomy for unruptured aneurysm,20)22) it is reasonable that there was no difference between clipping and endovascular coiling. However, our result was different from that of Otawara et al.18) who demonstrated lower cognitive function in the ruptured aneurysm group compared with the unruptured group. We think our limited patient number might have masked differences between groups.

We compared data from 59 patients who completed the K-MMSE more than once (Table 4). Although their pathologies and treatment modalities varied, 76.3% of (45 of 59) patients showed an improved K-MMSE score when they took the test again (mean: 11.9 months after the first test). However, it is unclear whether this change was due to cognitive function recovery or a practice effect due to repeated trials. Only one version of the questionnaire was available in our study, although other studies have utilized different versions administered in a random order to minimize practice effects.1) Because there was an interval of 11.9 months between the two tests, we think that practice effect were not a significant contributor. Similar cognitive improvements were also documented in several non-brain surgeries studies.14)15) Moller et al.14) indicated cognitive dysfunction in 25.8% and 9.9% of patients 1 week and 3 months after non-brain surgery, respectively. Newman et al.15) documented cognitive dysfunction in 36% and 24% of patients 6 weeks and 6 months after cardiac surgery, respectively. Abildstrom et al.1) reported that postoperative cognitive dysfunction is reversible in the majority of elderly patients after general surgery, such as major abdominal, non-cardiac thoracic, or orthopedic surgery with general anesthesia. Our results are different from those in non-brain surgery studies in two ways. First, our cognitive dysfunction rate was much higher, even after cognitive improvements. Although most patients showed an improved K-MMSE score and the number of the patients with cognitive dysfunction decreased from 41 to 18 (69.5% to 30.5%), many patients in our series still had cognitive dysfunction. The K-MMSE score did not show any linear correlation with the time interval between testing or with the time after treatment (p > 0.05). Therefore, we did not expect this improvement would continue. Second, our mean time from treatments to test was much longer than that observed in non-brain surgery. According to Moller et al.,14) dynamic cognitive improvement occurred within 3 months postoperatively. However, our study subjects showed similar improvement after a mean of 29.3 months postoperatively.

Our study has some limitations. First, we could not compensate all of the factors that might possibly bias K-MMSE scores; cognitive function is not only affected by pathologies or treatment modalities, it may also be influenced by many complicated socio-cultural factors, such as age, gender, education level, social class, marital status or income.3)5)7)17) This incomplete stratification and compensation stems from the uneven and relatively small number of patients, which is a drawback of recruiting various pathologies. Second, mortality and severe morbidity cases unable to visit the outpatient clinic were excluded from our study. We also excluded patients who refused to fill out our questionnaire. It is likely that there is something intrinsically different between people who volunteer for our study and those who do not.13) This is particularly true of neuropsychological studies, where personality traits might influence willingness to participate and test results.13) Thus, our result might have limitation in representing all kinds of neurosurgical patients.

Despite these limitations, however, our study contains unique information about the cognitive function of general neurosurgical patients, which can be used as a normative data for Koreans.